Can machine translation of literary texts fool stylometry?

This article uses standard authorship-attribution stylometry to tell machine translations made with DeepL and Google Translate from human translations. This is done using a Burrows-like distance measure procedure of cluster analysis, later visualized through network analysis. Using a corpus of Frenc...

Descripción completa

Detalles Bibliográficos
Publicado en:Digital Scholarship in the Humanities Vol. 40; no. 1; pp. 268 - 277
Autor principal: Rybicki, Jan
Formato: Artículo
Publicado: Oxford University Press / USA Apr2025
Materias:
Acceso en línea:Ver este registro en EBSCOhost
fields @attributes:
  recordID: 1
pdfLink:
plink: https://search.ebscohost.com/login.aspx?direct=true&db=hlh&AN=184296846&site=ehost-live
header:
  @attributes:
    shortDbName: hlh
    uiTerm: 184296846
    longDbName: Humanities International Complete
    uiTag: AN
  controlInfo:
    bkinfo:
    jinfo:
      jid:
        2055768X
        JEO9
      jtl: Digital Scholarship in the Humanities
      issn: 2055768X
      maglogo: N
    pubinfo:
      dt: Apr2025
      vid: 40
      iid: 1
      pid: 622
      pub: Oxford University Press / USA
    artinfo:
      ui:
        184296846
        10.1093/llc/fqaf010
      ppf: 268
      ppct: 9
      formats:
        fmt:
          – @attributes:
              type: T
          – @attributes:
              type: P
              size: 1.2MB
      tig:
        atl: Can machine translation of literary texts fool stylometry?
      aug:
        au: Rybicki, Jan
        affil: Department of Philology, Institute of English Studies, Jagiellonian University, 31-007 Kraków, Poland
      su:
        Machine translating
        Literature translations
        Warping machines
        Canon (Literature)
        Stylometry
      sug:
        subj:
          Machine translating
          Literature translations
          Warping machines
          Canon (Literature)
          Stylometry
      keyword:
        delta distance measure
        dynamic time warping distance
        French-to-English literary translation
        machine translation
        sentence length distribution
        stylometry
      ab: This article uses standard authorship-attribution stylometry to tell machine translations made with DeepL and Google Translate from human translations. This is done using a Burrows-like distance measure procedure of cluster analysis, later visualized through network analysis. Using a corpus of French literary classics translated into English by humans and machines as illustration, this article shows that, in most cases, translations of each text were very similar irrespective of the type of translator. Discrepancies in this respect were only found between translations of authors writing in very complex style (Proust). On the other hand, sentence length distribution compared with the Dynamic Time Warping Distance method was much more indicative of whether translations were made by humans or by machines.
      pubtype: Academic Journal
      doctype: Article
      src: R
    language: English
    refInfo:
    copyright:
      @attributes:
        flag: Y
      custom: © 2019 EADH: The European Association for Digital Humanities.
      item: Digital Scholarship in the Humanities
      holder: Oxford University Press / USA
      dt:
        @attributes:
          year: 2025
    holdings:
      @attributes:
        islocal: N